Several important clinical problems give rise to this proposal: 1) a given dose of antipsychotic produces widely varying plasma levels across individuals; 2) a therapeutic plasma level range must exist and a substantial proportion of relevant studies suggests that, at least for some commonly used antipsychotics, the upper limit falls within the range often achieved in clinical practice; 3) the lack of established response predictors and the resulting inclusion of patients unlikely to respond at any blood level may account for negative results in some plasma level/response studies; 4) inter-individual differences in plasma levels may be largely responsible for disagreement across response predictor studies. To address these problems the investigators propose to actively manipulate haloperidol plasma levels into and out of the supposed therapeutic range while monitoring subsequent clinical change. Concurrently, patients with various predictors of poor response will be alternately excluded to determine whether plasma level/response relationships are enhanced in the remaining patients. One-half of 110 patients with DSM-III schizophrenia will be assigned, on the basis of 24-hour pharmacokinetic studies, to doses projected to produce steady state haloperidol plasma levels of 8-18ng/ml; the other half will be assigned to doses selected to produce plasma levels of 25-35ng/ml. Plasma levels will then be determined at one and two and three weeks of fixed dose treatment. Patients who, at that point, have recovered according to study criteria will continue to take the same dose for an additional three weeks. Those who have not recovered at three weeks (estimated to the 75% of those who enter) will be randomly assigned a dose for the remainder of the study based on their week 2 blood levels. One-half will receive new doses designed to produce plasma levels in the alternate range while the other half will continue with their previous dose. Data from week 3 and week 6 will bear, in different ways, on the assumption that a clinically meaningful upper limit exists for the haloperidol plasma levels. Predictors of poor response to be tested will include predominately negative schizophrenia, markedly autistic premorbid adjustment, male sex, and an initially dysphoric response to haloperidol.